The Bitcoin vs Visa Electricity Consumption Fallacy

Last week, I got a Twitter alert about the following amusing reply of Andreas M. Antonopoulos to a yet another tweet complaining about the energy consumption of the Bitcoin network.

Lately, with the major crypto rally that we are seeing, articles about this topic have become extremely common and it seems that the only difference among them is to which country Bitcoin energy consumption is being compared with, whether is Nigeria , Iceland , Denmark or Ecuador.

While no one can argue that Bitcoin (and other altcoins) mining consumes a lot of electricity (in absolute numbers) given that you need to run a network of few hundreds or thousands of very powerful computers all the time, the right way to look at this problem is not about the total consumption but to compare how efficient is Bitcoin relative to the alternative traditional centralized systems that we are predominantly using today and that one day crypto might replace.

However, the only comparison that seems to always pop up everywhere is against VISA transaction costs which was included in the article that trigger the above tweet and in other articles as well. As expected, VISA looks way more efficient which adds to the rhetoric that Bitcoin is a very inefficient system and it is just a Ponzi scheme that is polluting the world. In my view, this comparison is flawed and it is not comparing apples to apples. Besides the fact that Bitcoin is not simply a piece of a payment network like VISA but a full currency system, VISA itself requires the banking system for its payment system to work so you need to actually include some of those costs there to make a meaningful comparison. So let’s look first at how VISA works.

How VISA actually works and who else is involved

Let’s take a close look at how a VISA transaction works when your cashier swipes your card (or these days, you wave your iPhone with Apple Pay). First, the data in your magnetic strip or phone is sent to something called a font-end processor that handles payment information on behalf of the merchant you used to pay and the bank where the merchant’s sales receipts are deposited. There are hundreds of front-end processors in the world. The bank where the merchant sales receipts are deposit is referred to as the acquiring bank. The front-end processors forward the information in the credit card to the relevant card association (VISA or others like MasterCard or American Express primarily) which then figures out which bank your card came from, the so-called issuing bank. Now, your payment information will be sent to a payment processor representing the issuing bank to validate that the information is correct and you have credit or balance. Finally, it is time for the transaction to flow back from the bank to the processor to the credit card association to the front-end processor to the merchant and to the acquiring bank for the transaction to actually occur. By now, your credit card information has crossed over several databases and servers out of which VISA is just one piece in the entire transaction chain.

So as we have seen above, just comparing VISA transaction costs with Bitcoin costs is not a fair comparison, we need to include the entire banking system electricity consumption since the foundation for VISA to work are the acquiring and issuing banks that participate in the transaction. And pretty much every single bank in the world is either an issuer or an acquiring bank or most of them in fact both. VISA is just one piece of a transaction while Bitcoin is the entire infrastructure needed to run a currency and a payment system.

Estimating banks electricity consumption

So let’s now estimate how much electricity the traditional banking system consumes. This is not an easy thing to do so the calculation below is just a rough estimate that tries to provide a lower bound by using only few sources of electricity for banks and using as conservative as possible numbers.

First let’s look at how many banks are out there. After doing a bit of research, it seems that no one knows exactly how many banks are in the world and you can get different estimates ranging from around 14,600 here to 25,000 here and even more than 60,000 quasi banks associations that are almost as rigorously regulated as banks. So for the purpose of this calculation we will just take the 30,000 number.

Now we need to estimate how much electricity banks consume. For the purpose of this comparison I will just include three values: server costs, branches costs and ATM costs. Of course, banks (and its employees), consume a lot more electricity from other sources but to make my point here this will be enough.

The next number we need to add to the mix are how many servers each bank is using to run their banking infrastructure. Pick a very conservative number of an average of 100 servers per bank (keep in mind banks need servers not just for the banking infrastructure but for the bank internal operations as well like ERPs, CRMs, accounting systems, Website, etc.). If a server in average consumes 400Wh and since it always on, this means that banks consume in total 800 Mwh.

Let’s add to the mix the electricity consumption of the branches. According to the World Bank there are 12.5 branches per 100,000 adults in the world so if the world population is 7.6 billion people and we have around 70% adults, this means a total of 665,000 branches. Only in the US they appear to be close to 100,000 branches and assuming US is around 15% or less of the entire banking system worldwide you get to around the same number.

Calculating branch consumption is more tricky since there are lots of things to take into account like size of the branch or number of employees as well as several things consuming electricity like lights, cooling, computers, monitors, etc. And they are not open 24 x 365 so after looking at a couple of articles, I have decided to settle for a conservative number 10 kwh per branch assuming an average branch has 10 light bulbs, two air conditioning units that are only use 20% of the time and 12 desktop computers running an average of 12 hours a day, 20 days a month through the year.

And finally, we need to include the ATMs networks that all banks use (that will also not needed in case bitcoin or other cryptocurrencies become the dominant currency and payment mechanism). According to the ATM Industry Association we were on the path to get 3MM ATMs by 2015 so given that we are in 2017 already I will just use that number as a lower bound for the total number. For small bank ATM machines, the average daily power consumption comes around 5Wh.

So total consumption for banks during a year only on those three metrics is around (I am rounding) 26Twh on servers, 58Twh on branches and 13Twh on ATMs for a total of close to a 100 Twh a year.

According to the article that trigger this discussion, Bitcoin annual Twh consumption is 28.67 , so currently more than 3 times more efficient than a very conservative calculation of the cost of the global banking system. Of course you will argue that the banking systems does more than handling a currency which is true but the difference is large enough that I do not think is that relevant. Even if only 30% of banks electricity consumption was the comparable part to Bitcoin, that will still make Bitcoin more efficient.

As a final comment, I believe that bitcoin will become more efficient in terms of electricity consumption moving forward (although it might continue growing its electricity consumption as an absolute value but so is energy consumption growing worldwide). The reasons are two fold.

First, as most of the miners move to the latest ASIC hardware like Bitmain’s Antminer S9, the cost per Th will become much more efficient. The S9 is 2.5 times from efficient in energy consumption than its predecesor the S7 and you can run a 14TH/s machine with only 1400 wats of energy. At the time of writing, the computers on the bitcoin network were doing 550,000 Terahashes per second. So if all of that compute was done with the most efficient hardware, you will need roughly 40K computers consuming 1400 wats each. Given that mining hardware is on all the time, you need to multiply this by 24 hours and 365 days given a total of around 500 Gigawats a year or as opposed of current estimates of 10,23 TWh to close to 30Twh yearly consumption, so 60 times more efficient. (note: this calculations are only taking into account the electricity costs of running the hardware and do not include the cost of the cooling but given how newer models like the S9 are also much more efficient in terms of heat dissipation and also reduce the amount of servers needed, adding this will only reinforce my point).

Second, the Bitcoin protocol itself is upgrading to improve its efficiency. We saw recently the SegWit activation and within the next few months we will see the adoption of the Lightning Network which effectively moves micropayments off the blockchain. Once this is done, Bitcoin will start to resemble more a clearing network with reduced fees and the electricity consumption per transaction will be decreased by several orders of magnitude.

So stop complaining about Bitcoin and start complaining about Xmas lights.